The High-Resolution Remote Sensing Image Change Detection Based on Differential Feature Fusion Siamese Pyramid Transformer Model
In recent years,deep learning (DL) techniques have made substantial breakthroughs in the field of remote sensing image change detection tasks. The Transformer-based method can better model global image information and has been studied and applied in the field of change detection. However,the existing transformer-based change detection methods only focus on the modeling of global information and ignore the importance of local details. To solve the above problems,this paper proposes a Siamese pyramid Transform-er change detection network based on differential feature fu-sion (DSTR). The model uses Siamese Pyramid Transform-er to build an encoder for multi-scale feature extraction,and a decoder using multilayer deconvolution to recover the original resolution of the feature map. At the same time,a differential feature fusion module is proposed,which uses the differential attention mechanism to fuse the difference information of dual-temporal features at different scales to improve the model change information extraction ability. Experiments on two public change detection datasets show that the proposed meth-od has achieved better detection results than the current rele-vant cutting-edge methods.
NETL
NSTL
万方数据
